Production web agents have a dirty secret. Most of their token budget goes to parsing bloated HTML — scripts, tracking pixels, ad divs — before the model sees the actual content. Web Speed, launched on Product Hunt on May 9 by Dominic Pi-Sunyer, attacks this head-on: a logic layer between the agent and the web… Continue reading

Prime Intellect flipped Lab from beta to GA on May 7. It’s a full-stack platform for training self-improving agents — define a task, write a harness, evaluate, run RL on the reward signal, inspect rollouts, deploy a LoRA adapter, serve inference. All inside one product. Nobody else has commercialized this loop end-to-end; until now you… Continue reading

ByteDance’s UI-TARS-desktop pulled 656 stars yesterday — sitting at 31.9k on GitHub, and everyone’s calling it the open-source answer to OpenAI’s Operator. What it actually is: two pieces. Agent TARS, a multimodal agent stack you run in a terminal, browser, or embed in your product. And UI-TARS Desktop, a native app that hands your computer… Continue reading

Airbyte spent eight years moving data into warehouses. Now they’re moving it into agents. Airbyte Agents is a managed Context Store that pre-replicates and pre-indexes data from Salesforce, Zendesk, Slack, Linear, Jira, Gong and 50+ other SaaS sources. Instead of an agent hitting a vendor MCP and burning thousands of tokens parsing raw API responses,… Continue reading

Anthropic shipped Dreaming on May 6, a research preview inside Claude Managed Agents. The pitch is blunt: stop letting your agents repeat the same mistake forever. What Dreaming actually does Between live sessions, the agent goes async and chews through its own past — transcripts plus the memory store. It pulls out patterns that survive… Continue reading